Reconfigurable fabric inspection system

Abstract

A method of identifying defects in a fabric can include obtaining an image of a fabric on a loom, extracting feature points within the image to generate an input image, processing the input image with a machine learning model, detecting one or more defects within the input image using the machine learning model, and providing, by the machine learning model, an indication of a defect in the input image.

What is claimed is: 
     
       1. A method of identifying defects in a fabric, the method comprising:
 obtaining an image of the fabric on a loom; 
 extracting feature points indicative of an intersection between a weft and a warp within the image to generate an input image, wherein the input image comprises a mapping of the feature points within the image, and wherein the input image has an area defined by a plurality of the feature points; 
 processing the input image with a first machine learning model; 
 detecting one or more defects within the input image using the first machine learning model; 
 providing, by the first machine learning model, an indication of a defect in the input image; and 
 providing a second machine leaning model, wherein the second machine learning model is configured to perform a further verification step of the one or more defects to confirm or deny the presence of the one or more defects of the detection from the first machine learning model, wherein the first machine learning model comprises a first neural network and the second machine learning model comprises a second neural network that is different from the first neural network, wherein the first neural network and the second neural network have different parameters, 
 wherein the identifying defects in the fabric is performed in real-time for a defect inspection that is to perform in-situ defect detection to pause faulty production of the fabric for correction, wherein the correction involves, when the further verification step of the one or more defects confirms the presence of the one or more defects, stopping the loom to correct a cause of the one or more defects. 
 
     
     
       2. The method of  claim 1 , wherein pre-processing comprises at least one of: performing a de-blurring process on the image, or normalizing the image. 
     
     
       3. The method of  claim 1 , wherein obtaining the image of the fabric on the loom comprises:
 using a contact image sensor (CIS) positioned adjacent the fabric on the loom to obtain the image of the fabric. 
 
     
     
       4. The method of  claim 3 , further comprising:
 controlling the CIS based on the image obtained from the CIS. 
 
     
     
       5. The method of  claim 4 , wherein controlling the CIS comprises:
 obtaining a positioning signal for the loom, wherein the positioning signal provides an indication of a position of the fabric on the loom; and 
 controlling a rate of imaging of the fabric on the loom based on the positioning signal. 
 
     
     
       6. The method of  claim 1 , wherein the machine learning model comprises a neural network, and wherein the neural network comprises a convolutional neural network or a spatial pyramid matching convolutional neural network. 
     
     
       7. The method of  claim 1 , wherein the defect comprises one or more of: warp streaks, reediness, weft bar, weft crack, thick and thin locations, weft loops, box marks, high incidence of warp breaks, weft breaks, shuttle traps, shuttle flying, smashes, bad selvedge, broken picks, bullet, half pick, broken ends, coarse ends, coarse pick, slough off, thick end and thick picks, double end, end out, fine end, jerk-in, knot, loom bar, loom barre, misdraw, mispick, reed mark, reed streak, set mark, shade bar, stop bar, tight end, stain, float, pin marks, contamination of fluff, harness misdraw, slubs, or combinations thereof. 
     
     
       8. The method of  claim 1 , further comprising:
 pausing an operation of the loom based on the indication of the defect in the input image. 
 
     
     
       9. A method of identifying defects in a fabric, the method comprising:
 obtaining an image of the fabric on a loom; 
 detecting one or more defects associated with a fault of the loom and/or a yarn problem within the image using a first machine learning model; 
 sending the image to a system controller in response to detecting the one or more defects within the image; and 
 providing a second machine leaning model, wherein the second machine learning model is configured to perform a further verification step of the one or more defects to confirm or deny the presence of the one or more defects of the detection from the first machine learning model, wherein the first machine learning model comprises a first neural network and the second machine learning model comprises a second neural network that is different from the first neural network, wherein the first neural network and the second neural network have different parameters, 
 generating, by the system controller, an output signal in response to the verification of-the presence of the one or more defects within the image, 
 wherein the method of identifying defects in the fabric is performed in real-time that is to perform in-situ defect detection to pause faulty production of the fabric for correction, wherein the correction involves, when the further verification step of the one or more defects confirms the presence of the one or more defects, stopping the loom to correct a cause of the one or more defects. 
 
     
     
       10. The method of  claim 9 , wherein pre-processing comprises at least one of: performing a de-blurring process on the image, or normalizing the image. 
     
     
       11. The method of  claim 9 , wherein obtaining the image of the fabric on the loom comprises:
 using a contact image sensor (CIS) positioned adjacent the fabric on the loom to obtain the image of the fabric. 
 
     
     
       12. The method of  claim 9 , wherein the first neural network and the second neural network comprise different parameters. 
     
     
       13. The method of  claim 9 , wherein the defect comprises one or more of: warp streaks, reediness, weft bar, weft crack, thick and thin locations, weft loops, box marks, high incidence of warp breaks, weft breaks, shuttle traps, shuttle flying, smashes, bad selvedge, broken picks, bullet, half pick, broken ends, coarse ends, coarse pick, slough off, thick end and thick picks, double end, end out, fine end, jerk-in, knot, loom bar, loom barre, misdraw, mispick, reed mark, reed streak, set mark, shade bar, stop bar, tight end, stain, float, pin marks, contamination of fluff, harness misdraw, slubs, or combinations thereof. 
     
     
       14. The method of  claim 9 , wherein the presence of the one or more defects within the image are verified via the second machine learning model after the one or more defects are detected via the first machine learning model. 
     
     
       15. A system for identifying defects in a fabric in real-time, the system comprising:
 a contact image sensor (CIS) configured to be coupled to a loom and image the fabric on the loom; and 
 a CIS controller in signal communication with the CIS, wherein the CIS controller comprises a processor, wherein the CIS controller is configured to: 
 receive an image of the fabric on the loom; 
 process the image with a first machine learning model; 
 identify one or more defects associated with a fault of the loom and/or a yarn problem in the fabric using the first machine learning model; and provide an output indicative of the presence of the one or more defects in the fabric; and 
 a system controller, wherein the system controller comprises a processor and is in signal communication with the CIS controller and a memory storing instructions for the system, wherein the system controller is configured to:
 receive, from the CIS controller, the output indicative of the presence of the one or more defects; 
 receive, from the CIS controller, the image of the fabric on the loom; 
 process-the image of the fabric on the loom using a second machine learning model via a further verification step of the one or more defects to confirm or deny the presence of the one or more defects of the detection from the first machine learning model; 
 verify the presence of the one or more defects in the fabric via the second machine learning model; and 
 provide a second output indicative of the presence of the one or more defects in the fabric, 
 
 wherein the first machine learning model comprises a first neural network and the second machine learning model comprises a second neural network that is different from the first neural network, wherein the first neural network and the second neural network have different parameters, 
 wherein the identification of defects in the fabric is performed in real-time that is to perform in-situ defect detection to pause faulty production of the fabric for correction, wherein the correction involves, when the further verification step of the one or more defects confirms the presence of the one or more defects, stopping the loom to correct a cause of the one or more defects. 
 
     
     
       16. The system of  claim 15 , wherein the CIS comprises a light source and an imaging sensor. 
     
     
       17. The system of  claim 15 , wherein the CIS comprises a plurality of CIS devices arranged in an array, wherein the array is as wide as or wider than a width of the fabric, wherein at least one edge of each CIS device of the plurality of CIS devices overlaps with an adjacent CIS device of the plurality of CIS devices within the array. 
     
     
       18. The system of  claim 15 , further comprising:
 pre-processing the image prior to generating the input image, wherein the pre-processing comprises at least one of: performing a de-blurring process on the image, or normalizing the image. 
 
     
     
       19. The system of  claim 15 , further comprising:
 one or more mobile devices in signal communication with the system controller, wherein the one or more mobile devices are configured to receive the second output from the system controller. 
 
     
     
       20. The system of  claim 19 , wherein the one or more mobile devices are configured to display the image and receive an input from an operator to verify a presence of the defect in the image. 
     
     
       21. The system of  claim 15 , wherein the system controller is configured such that the presence of the one or more defects within the image are verified via the second machine learning model after the one or more defects are detected via the first machine learning model.